Analysis of function in the absence of extant functional homologues: a case study using mesotheriid notoungulates (Mammalia)

Page created by Howard Strickland
 
CONTINUE READING
Paleobiology, 33(2), 2007, pp. 227–247

Analysis of function in the absence of extant functional
homologues: a case study using mesotheriid notoungulates
(Mammalia)

Bruce J. Shockey, Darin A. Croft, and Federico Anaya

Abstract.—We use two approaches to test hypotheses regarding function in a group of extinct mam-
mals (Family Mesotheriidae, Order Notoungulata) that lack any close extant relatives: a principle-
derived paradigm method and empirically derived analog method. Metric and discrete morpho-
logical traits of mesotheriid postcranial elements are found to be consistent with the morphology
predicted by a modified version of Hildebrand’s paradigm for scratch diggers. Ratios of in-force
to out-force lever arms based on skeletal elements indicate that the mesotheriids examined had
limbs modified for high out-forces (i.e., they were ‘‘low geared’’), consistent with the digging hy-
pothesis. Other mesotheriid characters, such as cleft ungual phalanges, a curved olecranon, and a
highly modified pelvis (with extra vertebrae incorporated into the sacrum and fusion between the
ischium and the axial skeleton) are regarded as being functionally significant for digging and also
occur in a variety of extant diggers. Analog methods indicate that mesotheriids share numerous
traits common to a variety of extant diggers. Principal component analyses of postcranial elements
indicate that mesotheriids consistently share morphometric space with larger extant fossorial
mammals: aardvark, anteaters, wombats, and badger. Likewise, discriminant function analyses cat-
egorized mesotheriids as fossorial, though imperfectly analogous to the extant diggers analyzed.
Thus, both theory-driven and empirically derived methods of estimating function in these extinct
taxa support a digging hypothesis for the mesotheriids examined. Adaptations for digging in both
the forelimb and sacropelvic functional complexes of mesotheriids provide independent support
for the fossorial hypothesis.

Bruce J. Shockey. Department of Vertebrate Paleontology, American Museum of Natural History, New York,
  New York 10024. E-mail: bshockey@amnh.org
Darin A. Croft. Department of Anatomy, Case School of Medicine, Case Western Reserve University, Cleve-
  land, Ohio 44106-4930. E-mail: dcroft@case.edu
Federico Anaya Daza. Facultad de Ingenierı́a Geológica, Universidad Autónoma ‘‘Tomás Frı́as,’’ Potosı́, Bo-
  livia

Accepted:    27 September 2006

                      Introduction                               rived from the paradigm method. Theory-
                                                                 driven and empirically derived hypotheses
   Functional reconstruction of life in the past
                                                                 serve as at least quasi-independent tests for
is challenging, especially for extinct taxa that
                                                                 one another (Rudwick 1964). Likewise, func-
lack extant functional homologues. To be con-
fident in functional interpretations for such                    tional hypotheses based on metric variables
taxa, any proposed hypothesis must somehow                       (lengths of elements, and ratios of these
be tested. In this work, we provide an example                   lengths) may be compared with those based
that uses principles from functional morphol-                    upon discrete (presence-absence) morpholog-
ogy to predict the morphology for a given be-                    ical characters. When both types of data are
havioral complex and then test that prediction                   compatible with the hypothetical function,
by empirically derived means. In this case, a                    then confidence in the hypothesis is much
modified version of the ‘‘scratch digging’’                      greater.
paradigm (sensu Hildebrand 1974, 1985) is                           Our example here is from the extinct, en-
used to predict morphological patterns in a                      demic, South American notoungulates (Order
mammal that habitually digs (see ‘‘Meth-                         Notoungulata Roth, 1904) and includes three
ods’’). Comparing morphologies of the puta-                      taxa in the family Mesotheriidae Alston, 1876.
tive fossorial fossil taxa and a variety of un-                  Mesotheriids were rabbit- to mostly sheep-
related extant mammals in order to find close                    sized notoungulates distinguished by their
functional analogs tests the hypothesis de-                      ever-growing, gliriform, chisel-like incisors,
䉷 2007 The Paleontological Society. All rights reserved.                                      0094-8373/07/3302-0004/$1.00
228                                        SHOCKEY ET AL.

and hypsodont to hypselodont cheek teeth.           theres with toxodonts and various other ex-
Two subfamilies are recognized, Trachytheri-        tinct, endemic, South American ungulates in
inae Ameghino, 1894, and Mesotheriinae Al-          his Order Notoungulata, a scheme that has
ston, 1876. Trachytheriines include basal           been stable (with minor revisions of included
forms that may constitute a paraphyletic as-        taxa) for a hundred years.
semblage of taxa (Reguero and Castro 2004);            Various hypotheses regarding the biology
they occur in faunas referable to the late Eo-      of mesotheriids have been proposed. From his
cene (Divisaderan South American Land               comparison of the forelimb of Mesotherium
Mammal ‘‘Age’’ [SALMA]; but see Cerdeño et         with those of beavers (Castor), river otters (Lu-
al. 2006 for an alternative interpretation of the   tra), and seals (Phocidae), Serres (1867) inter-
Divisaderan trachytheriine) to late Oligocene       preted Mesotherium as an aquatic animal.
(Deseadan SALMA). Mesotheriines constitute          Loomis (1914) characterized the typotheres
a monophyletic group recognized by numer-           (including Trachytherus) of the Deseado fauna
ous synapomorphies (see Croft et al. 2004),         (Deseadan SALMA) as ‘‘running and hopping
the most conspicuous being the loss of the up-      animals (p. 53)’’ and implicitly characterized
per second incisor through second upper pre-        Trachytherus, by way of its ever-growing inci-
molar and lower third incisor to third pre-         sors, as ‘‘gnawing the bark and eating the
molar (I2–P2/i3–p3) and the hypselodont             twigs and leaves of bushes (p. 61).’’ Heidi Sy-
(ever-growing) condition of all teeth. The ear-     dow (1988), noted many morphological char-
liest well-preserved mesotheriines are from         acteristics of Trachytherus that are associated
the early Miocene (Santacrucian SALMA) of           with fossoriality in modern mammals. She in-
northern Chile (Croft et al. 2004) and the clade    terpreted Trachytherus as being a ‘‘scratch dig-
persisted until the Pleistocene (see below).        ger’’ (sensu Hildebrand 1974 and 1985), an in-
   The Pleistocene Mesotherium Serres 1867          terpretation tested and supported by the pres-
was the first known mesotheriid and for sev-        ent study. Apparently because of the similar-
eral decades was widely known by its junior         ities between mesotheriids and the semi-aquatic
synonym Typotherium Gervais, 1867 (see              rodent Hydrochaeris (Family Hydrochaeridae),
Simpson 1940 for details of this somewhat           Bond et al. (1995) used the capybara as a mod-
complicated taxonomic problem). For several         el for Pseudotypotherium and Mesotherium and
decades, Typotherium served as the type genus       suggested that these mesotheriids were ‘‘cur-
for the family Typotheriidae Lydekker, 1884         sorial y semiacuático’’ (Bond et al. 1995: p. 264)
(⫽ Mesotheriidae Alston, 1876) and it remains       and probably grazers.
the namesake of the currently recognized sub-          In this study, we attempt to gain some un-
order Typotheria Zittel, 1892.                      derstanding of mesotheriid biology. Because
                                                    mesotheriids—indeed all notoungulates—are
‘‘. . . le Mesotherium est réellement un animal    extinct, and because the interordinal relation-
paradoxal . . . .’’—Serres 1867: p. 6.              ships of notoungulates are unresolved, it is
   Serres (1867) christened the beast Mesother-     impossible to gain any useful functional in-
ium (‘‘middle beast’’) because he believed the      sights by using an extant phylogenetic bracket
animal was an intermediate between two dis-         (sensu Bryant and Russell 1992; Witmer 1995):
tinct orders of mammals. Noting an animal           such a bracket would be so large as to be
with rodent-like incisors and an ungulate-like      meaningless for our purposes. Therefore, we
body, but with a clavicle (absent in most un-       rely upon a principle-to-practice approach
gulates), he suggested that Mesotherium was in      known as the ‘‘paradigm method’’ (Rudwick
transition from a rodent to a ‘‘pachyderm’’         1964; see also Gould 1970) and we test the re-
(Serres 1867: p. 7). The odd suite of characters    sults via empirical methods (i.e., we compare
of Mesotherium and other typotheres inspired        the morphology of the extinct taxa with extant
a variety of hypotheses regarding their phy-        taxa whose behavior is known).
logenetic relationships, including being relat-        For the principle-based paradigm method,
ed to prosimian primates (Ameghino 1891,            we use a modified form of Hildebrand’s
1906). Ultimately, Roth (1903) united typo-         scratch-digging paradigm (1974, 1985) as an a
MESOTHERIID MORPHOLOGY                                        229

priori ‘‘prediction’’ of the morphological char-    paradigm the observation that the pubic sym-
acter states to be found in an animal that ex-      physis is weak or absent in most diggers. No
cavates by ‘‘scratch digging’’ (i.e., extending     functional principle predicted that such
the forefeet to the substrate, breaking the sub-    should occur in diggers and, indeed, its func-
strate, and pulling the material under the          tion is not understood. Because the function is
body by flexion of digits and wrist, elbow ex-      unknown and is not predicted from principle,
tension, and forelimb retraction). Scratch dig-     we do not include it in our version of the par-
gers must be capable of directing great force       adigm. Conversely, from lever mechanics we
against the substrate in order to excavate.         would predict a high greater tubercle of the
Thus, their forelimbs must be able to produce       humerus, because it would provide a mechan-
large out-forces (Fo ), which may be expressed      ical advantage (high Li) for protracting the hu-
as:                                                 merus. We therefore include this feature in our
                                                    version of the scratch-digging paradigm even
                 Fo ⫽ Fi Li/Lo                (1)
                                                    though it is not common among living diggers
where Fi is the in-force, Li the in-lever, and Lo   (see Larson and Stern 1989 for discussion of
the out-lever. By inspection of the equation        greater tubercle functions).
above, one can see that there are three ways to        In this study, we examine the postcranial re-
increase Fo: (1) increase Fi by increasing the      mains (and to a lesser extent, cranial remains)
cross-sectional area of the muscle(s) that act      of mesotheriids of both subfamilies, Trachy-
upon the lever; (2) increase Li by increasing the   theriinae and Mesotheriinae, and compare
distance between the point of muscle insertion      them with the scratch-digging paradigm (de-
and the point of rotation (generally requiring      tailed below). As a quasi-independent confir-
lengthening the bone acting as Li), or (3) de-      mation (Rudwick 1964), we also compare me-
crease Lo by shortening the bone that acts as       sotheriid morphologies with those of extant
the out-lever. All three of these approaches are    scratch diggers and other fossorial animals.
seen in extant scratch diggers. For example,        The scratch-digging hypothesis should be re-
the armadillo Dasypus has huge triceps brachii      garded as falsified should the extinct taxa
muscles with extensive areas of origin and in-      have fewer modifications of the postcranial
sertion (large Fi), a relatively long olecranon     skeleton than extant scratch diggers. Of
(large Li), and a short ulnar shaft (small Lo)      course, such a falsification would not imply
(Hildebrand 1985).                                  that the animals did not dig (dogs dig, but
   Evaluating such features in fossil specimens     would fail this test), but rather that the fos-
can be problematic; direct measures of cross-       sorial hypothesis would not be robust and that
sectional muscle areas are almost never avail-      digging may not have been a specialization for
able for fossils, and osteological evidence re-     the animal. To fail to falsify the hypothesis re-
garding musculature may be unreliable (Bry-         quires that the fossils show a greater number
ant and Seymour 1990). Many of the muscu-           of and/or more extreme adaptations for dig-
loskeletal adaptations of fossorial animals are     ging than some extant diggers. This method,
so extreme, however, that they frequently           of course, will generate more false negatives
leave conspicuous muscle scars, crests, and tu-     than false positives, but we desire difficult
berosities on the bones, thus often rendering       tests for functional hypotheses regarding ex-
general interpretations of major muscles un-        tinct animals lacking close extant relatives. We
ambiguous (see examples below and Fig. 1).          must be content with our ignorance about life
Also, evidence of the levers of the musculo-        in the past, when appropriate, but robust
skeletal system, the bones themselves, is com-      functional hypotheses need not be ignored.
monly available to vertebrate paleontologists.
   In its ideal form, the paradigm method de-                  Materials and Methods
rives a predicted morphology from biome-              Abbreviations. Institutional abbreviations
chanical principles. In some cases, however, it     are as follows: CMNH, Cleveland Museum of
may be more empirically based. For example,         Natural History, Ohio; FLMNH, Florida Mu-
Hildebrand (1985) included in his digging           seum of Natural History, Gainesville; UF, Ver-
230                                            SHOCKEY ET AL.

FIGURE 1. Left forelimb elements of selected fossorial mammals and the generalized ambulatory marsupial Di-
delphis compared with those of the mesotheriid Trachytherus. Humeri are illustrated in anterior view above, ulnae
in anterior view below. Elements are from the following specimens: Didelphis virginiana (UFm 21682); Trachytherus
spegazzinianus (UF 91933); Orycteropus afer (CMNH 18504); Taxidea taxus (UFm 6734); Tamandua tetradactyla (UFm
10119); Vombatus ursinus (CMNH 18946). Scale bars, 3 cm.
MESOTHERIID MORPHOLOGY                                        231

tebrate Paleontology Division of FLMNH, and          dylar processes. Fossorial animals almost in-
UFm, Mammalogy Division of the FLMNH,                variably have DHW ⬎30 owing to an enlarged
both of the University of Florida, Gainesville;      medial epicondylar process for the attachment
MNHN-Bol, Museo Nacional de Historia Nat-            of enlarged wrist and digit flexor muscles. The
ural, La Paz, Bolivia; MUSM, Museo de His-           golden mole, Amblysomus, is an extreme ex-
toria Natural, Universidad Nacional Mayor de         ample, having its medial epicondylar process
San Marcos, Lima, Peru.                              extended to such a degree that its distal width
   Other abbreviations are as follows: SALMA,        is nearly the same as the entire humeral length
South American Land Mammal ‘‘Age’’; PCA,             (DHW ⫽ 98) (Hildebrand 1985). We note here
principal components analysis; DFA, discrim-         that high DHW values in diggers can also re-
inant function analysis. Abbreviations for le-       sult from relative shortening of the humerus;
ver mechanics are Fo, out-force, Fi, in-force, Lo,   a long humerus would be mechanically dis-
out-lever, Li, in-lever, Vo, out-velocity, and Vi,   advantageous, reducing Fo while increasing
in-velocity. Indices (defined below) are BI,         Lo. Cursors have low DHW values.
brachial index; CI, crural index; DHW, distal           Deltopectoral Crest Index (⌬PC) is 100 ⫻
humeral width index; ⌬PC, deltopectoral              (length of the deltopectoral crest/length of
crest index; MtFI, metatarsal index; OI, olec-       humerus). This is examined as a means of es-
ranon index.                                         timating the relative out-force for humeral
   Indices. Various ratios useful for estimat-       protraction.
ing functional abilities were calculated from           Olecranon Index (OI) is 100 ⫻ (olecranon
the specimens we examined and were supple-           length/ulnar shaft length); the length of the
mented by published data for Mesotherium             olecranon is measured from the tip of the olec-
(Serres 1867) and other taxa (e.g., Coombs           ranon to the midpoint of the trochlear notch
1983; Hildebrand 1985; Van Valkenburgh               (sensu Hildebrand 1985 and Van Valkenburgh
1987; Garland and Janis 1993). Such indices          1987) and the ulnar shaft is measured from
taken by themselves have limited value in pre-       the midpoint of the trochlear notch to the dis-
dicting function, but are useful when consid-        tal tip of the ulna, ignoring any stylar process.
ered in the context of an animal’s phylogenetic      OI here is not directly comparable to that of
relationships and body size (see Garland and         Coombs 1983, which offers greater precision
Janis 1993). Indices used in this study are ra-      but does not estimate the ratio of the levers.
tios expressed as percentages (i.e., ⫻100).          Fossorial mammals have olecranon indices
They are calculated for mesotheriids in              ⬎25, usually ⬎30 (Hildebrand 1985; Van Val-
Appendix 1 (http://dx.doi.org/10.1666/pbio           kenburgh 1987), and at times much higher
05052.s1) and are defined as follows:                (e.g., 95 in the giant armadillo, Priodontes, ac-
   Brachial Index (BI) is the ratio of the length    cording to Hildebrand’s [1985] data).
of the radius to the length of the humerus              Crural Index: (CI) is 100 ⫻ (tibia length/fe-
(measured from the head to the trochlea, ne-         mur length). Cursorial and saltatory animals
glecting the height of the greater tuberosity)       tend to have elongated distal limb elements
times 100 (100 ⫻ [radius length/humerus              and almost invariably have CI values ⬎100,
length]). Mammals specialized for digging al-        whereas diggers and many climbers have CI
most invariably have a radius that is shorter        values ⬍100 (Hildebrand 1974).
than the humerus (BI ⬍ 100; Hildebrand 1974;            Metatarsal/Femur Index (MtFI) is 100 ⫻
Coombs 1983). Exceptions are few and include         (metatarsal III length/femur length). Fossorial
diggers that also are good runners (e.g., some       animals tend to have MtFI values ⬍30, cur-
rabbits [e.g., Lepus]; see Coombs 1983) and the      sorial carnivores have MtFI values of 35–50,
terrestrial anteater Myrmecophaga (Coombs            and most cursorial ungulates—though no
1983; Taylor 1985).                                  faster than cursorial carnivores—have ex-
   Distal Humerus Width (DHW) is 100 ⫻ (dis-         treme MtFI values, 50–150 (Garland and Janis
tal humeral width/humeral length); distal            1993).
humeral width is measured over the trochlea             Specimen Information. We examined the
and includes the distal portions of the epicon-      skeletal remains of mesotheriids from two dis-
232                                       SHOCKEY ET AL.

tantly related genera; one is a trachytheriine      phaga), animals that excavate social insects
(Trachytherus spegazzinianus) and the other is a    from a variety of substrates. We also sampled
mesotheriine (Plesiotypotherium sp.). The ma-       taxa that exhibit other locomotor functions, in-
terial studied is as follows:                       cluding those that are known to be semiaquat-
   Trachytherus spegazzinianus (late Oligocene,     ic, cursorial, or generalized. The taxa consid-
Deseadan SALMA, Salla, Bolivia): UF 91933,          ered (specimens and references) are listed in
skull, mandibles, and associated postcrania         Appendix 1.
including left and right forelimbs, mostly             Paradigm Method. We use the scratch-dig-
complete; UF 90960, cranium, left scapula, left     ging paradigm of Hildebrand (1974, 1985),
humerus, left and right ulnae, left radius, left    though in a modified form to exclude his em-
metacarpals (Mc) III, IV, and V, and right Mc       pirically derived characters. Our version of
III and IV, several vertebrae, left innominate,     the scratch-digging paradigm (1–5 below)
left and right femora and tibiae, and proximal      lists the biomechanical demands of scratch
tarsus (UF 91933 and 90960 were described in        digging and the morphological characters (in
Sydow 1988); UF 172437, right astragalus, left      italics) related to meeting those demands. The
tibia; UF 173257, right astragalus; UF 172514,      major requirement of scratch digging is that
right calcaneum; MNHN-Bol field #94-02, left        the forelimb generates significant out-forces
astragalus (Shockey 1997a: Fig. 6.6a).              (Fo) against the substrate. Morphological fea-
   Plesiotypotherium sp., ?Pliocene, Casira, Bo-    tures (sensu Bock and von Wahlert 1965) that
livia: MNHN-Bol-3724, partial skeleton in-          would facilitate the generation of Fo are as-
cluding partial cranium, mandible, left hu-         sessed in terms of putative functional com-
merus (in articulation with the ulna and ra-        plexes or faculties (1–6). These include (1)
dius, including a radial sesamoid), partial         shoulder (i.e., glenohumeral) joint for humeral
left? manus, several vertebrae, and a partial       retraction and (2) shoulder joint for humeral
pelvis (including extra fused sacral [fused         extension; (3) elbow joint; (4) wrist and digits.
caudal] vertebrae, partial left ilium, left pubis   The earth opposes these significant out-forces,
and ischium). This specimen was collected at        which in turn pass back into the body and
Casira by one of us (F.A.D.) and Pierre-An-         need to be opposed by (5) reinforcing/sup-
toine Saint-André, formerly of the Muséum         portive structures in the hands. As the forces
National d’Histoire Naturelle, Paris.               pass through the body, the digger must be
   Postcranial data for Mesotherium cristatum       firmly ‘‘planted’’ in the substrate via the pos-
were obtained from the descriptions of Serres       terior limbs (or, in some cases, the tail); thus,
(1867) and the plates for this work, which were     there should be (6) reinforcing/supportive
published later by Gervais (1869). Other de-        structures in the posterior appendicular skel-
scriptions of Tertiary notoungulates used for       eton.
comparison included those of the isotemnid             Some of these characters are discrete and do
Thomashuxleya (Simpson 1967); the notohippid        not require quantification (i.e., they may be as-
Eurygenium pacegnum (Shockey 1997a,b); the          sessed as present or absent) whereas others
leontiniid Scarrittia canquelensis (Chaffee         are relative and are quantified via the mea-
1952); the toxodontids Nesodon and Adinother-       sures indicated in Appendix 1 and defined
ium (Scott 1912); and the typotheres Protypo-       here:
therium, Interatherium, and Hegetotherium (Sin-
                                                       1. High Fo generated through the shoulder
clair 1909).
                                                    joint for humeral retraction in order to pull
   We also examined osteological specimens
                                                    substrate under the body. This high Fo may be
and reviewed the literature regarding a vari-
                                                    generated via high Fi as evidenced by:
ety of extant fossorial animals and behavior-
ally and functionally similar ‘‘tearing’’ ani-        Acromion enlarged (extends to or beyond the
mals (sensu Coombs 1983), also known as                 glenoid) and/or spine of scapula raised to
‘‘hook and pull’’ diggers (sensu Hildebrand             accommodate a well-developed acrom-
1985); this latter group includes the myrme-            iodeltoid muscle;
cophagid anteaters (e.g., Tamandua, Myrmeco-          Deltopectoral crests enlarged and distinctive to
MESOTHERIID MORPHOLOGY                                                233

    provide significant surface area for en-
    larged deltoid and pectoralis muscles;
  Posterior angle of scapula extended, forming a
    distinct postscapular fossa for the origin
    of an enlarged teres major. (Also increas-
    es Li by way of its more distant location
    from the shoulder joint.)
  High Fo for humeral retraction may also be
generated through the shoulder joint by way
of an increased Li and would be evidenced by:
  Deltopectoral crests positioned distally (mid-
    way down the shaft or further) for inser-
    tion of deltoid and pectoralis muscles far
    from the shoulder joint.
  2. High Fo for humeral protraction in order
to break the substrate or to inhibit (oppose)
the actions of retraction (noted above). This
high Fo may be generated by high Fi for hu-
meral protraction as evidenced by:
  Scapular spine high (height of spine above
    blade ⬎ dorsoventral diameter of glen-
                                                   FIGURE 2. Forelimb elements of Trachytherus spegazzin-
    oid) to accommodate massive supraspi-          ianus (UF 91933) of Salla, Bolivia (late Oligocene). A,
    natus and infraspinatus muscles.               Left humerus in anterior (left) and lateral (right) views.
                                                   B, Left manus (dorsal view). C, Left ulna with pisiform
  High Fo for humeral protraction may also be      in anterior (above) and medial (below) views, distal to
accomplished by way of an increased Li as in-      right. (Scale bar applies to all.)
dicated by:
                                                       provide greater surface area for attach-
  Greater tuberosity higher than humeral head to
                                                       ment of enlarged pronator teres and wrist
    increase Li for supraspinatus.
                                                       flexors, thus generating increased Fi;
   3. High Fo generated through the elbow for        Olecranon inflected medially for enlarged ori-
flexion necessary for substrate removal as in-         gin of wrist and digit flexors (Figs. 1, 2);
dicated by:                                          Pisiform elongate and robust to increase Li for
                                                       the flexor carpi ulnaris (e.g., the badger,
  Olecranon long (OI ⬎ 30) to provide greater
                                                       Taxidea [Hildebrand and Goslow 2001]);
    moment arm (Li) for extension and great-
                                                     Metacarpals short, creating small Lo, thereby
    er insertion area for triceps muscles and
                                                       increasing Fo.
    other elbow extensors (anconeus, tensor
    fasciae antebrachii);                            5. Resistance to ground forces. As the large
  Radius short (BI ⬍ 100), shorter than humer-     Fo generated by the forelimb acts upon the
    us, thus decreasing Lo and thereby in-         substrate, the substrate will have an equal and
    creasing (Fo).                                 opposite force upon the animal. Adaptations
                                                   in response to these forces of the earth will be
  4. High Fo generated through the wrist (car-
                                                   evident by:
pal), metacarpophalangeal, and interphalan-
geal joints:                                         Ungual phalanges fissured to secure claw, nail,
  High Fo for digit and wrist flexion is nec-          or hoof to ungual phalanx;
essary to break and remove the substrate and         Bony dorsal ‘‘stops’’ on phalanges to prevent
will be evident by:                                    overextension of digits.
  Entepicondylar process large (DHW ⬎ 30) to         6. Transfer of force through body. Forces
234                                       SHOCKEY ET AL.

transferred to the posterior body from the           categories, we acknowledge that these ani-
earth via the forelimb during excavation are         mals have locomotor repertoires that may in-
opposed by:                                          clude several types of behaviors (see ‘‘Discus-
                                                     sion’’). The multivariate analyses were execut-
  Supernumerary fused sacral vertebrae (or
                                                     ed using SPSS 11.0.2 (SPSS Inc.) on an Apple
     ‘‘pseudosacrals’’ [e.g., fused caudal ver-
                                                     PowerBook G4 computer.
     tebrae]),
                                                        Qualitative and discrete morphological fea-
  Additional sacropelvic contacts to reinforce the
                                                     tures of functional significance were also not-
     pelvic girdle and/or
                                                     ed and compared among the extant taxa and
  ‘‘Low-geared’’ hind limbs (CI ⬍ 100 and MtFI
                                                     the mesotheriids. Examples of such characters
     ⬍ 40)
                                                     relevant to scratch digging in mesotheriids in-
   Analog Method. We tested the results of the       clude bifid ungual phalanges, curved olecra-
paradigm method by comparing the mor-                non, supernumerary fusion of vertebrae at the
phologies of mesotheriids (Trachytherus and          sacrum, and sacroischial fusion.
Mesotherium) with those of 30 species of mam-
mals of known locomotor function including                Comparative Functional Anatomy
scratch diggers, runners, frequent swimmers             This section is devoted primarily to an ele-
(i.e., semiaquatic mammals), and generalists         ment-by-element description of the forelimb,
(Appendix 1). The notohippid Eurygenium pa-          pelvic girdle, and hind limb of Trachytherus
cegnum Shockey 1997b was also included in            spegazzinianus and Plesiotypotherium sp.; these
the analyses as a putative example of an un-         are compared to Serres’s description of Me-
specialized notoungulate. Nine measure-              sotherium cristatum (Serres 1867) and to a va-
ments were recorded for each specimen: (1)           riety of other taxa including notoungulates
humeral shaft length (head to trochlea, ne-          and extant taxa whose general ecomorpholo-
glecting any extension of the greater tubercle       gy is known. Metric data are provided in Ap-
beyond the head), (2) distal humeral width           pendix 1.
(across and including the condyles), (3) length         Scapula. UF 90960 preserves most of the
of deltopectoral crests (from the humeral head       left scapula of Trachytherus spegazzinianus. The
to the point where they meet and terminate),         spine is high, relatively higher than the spine
(4) olecranon length (tip to midpoint of troch-      of Protypotherium (see Sinclair 1909), rising to
lear notch), (5) ulnar shaft length (midpoint of     a height (22.8 mm) that exceeds the diameter
trochlear notch to distal end of ulna), (6) total    of the glenoid fossa (20.6 mm). The acromion
length of radius, (7) femoral shaft length           is fairly well developed, reaching a point
(head to distal end, neglecting any extension        above, but not beyond, the coracoid process.
of greater trochanter), (8) tibial shaft length,     Its length is much greater than that of Proty-
and (9) metatarsal (Mt) III length.                  potherium and the basal toxodontids Nesodon
   The nine measurements described above             and Adinotherium (see Scott 1912); although
were analyzed using principal components             the tip is broken, it does not appear to be ex-
analysis (PCA) and discriminant function             tended to the degree exhibited by many xe-
analysis (DFA). The PCA was performed in or-         narthrans. There does not appear to have been
der to assess variation among the taxa exam-         a metacromion process or a ventrally directed
ined and included nine appendicular vari-            spinal process as in Nesodon or Adinotherium
ables. As a means of objectively assigning the       (see Scott 1912). Damage to the scapular blade
fossil taxa to functional groups based on limb       does not permit an assessment of the presence
bone measurements, we also performed a               or absence of an extension of the posterior an-
DFA using the same nine variables; the prior         gle.
probabilities of the four locomotor categories          Serres (1867) described the spine of the
(generalized, fossorial, cursorial, semiaquatic)     scapula of Mesotherium as being quite high
were considered equal, and the fossil taxa           above the scapular body. The figure of the
were coded as unknown. Although we clas-             scapula (Gervais 1869: Plate XXV.1) shows a
sified the extant taxa in discrete locomotor         well-developed, single metacromion; that of
MESOTHERIID MORPHOLOGY                                                235

the toxodontid notoungulate Nesodon has two
metacromia (Scott 1912). There is no signifi-
cant hypertrophy of the posterior angle.
   Humerus. UF 90960 and UF 91933 include
complete humeri of Trachytherus spegazzini-
anus. The most conspicuous features of the hu-
merus are its robust nature and well-defined
crests and muscle scars (Figs. 1, 2). The crests
for the deltoid and pectoralis muscles are pro-
nounced and extend two-thirds the length of
the element (Appendix 2, http://dx.doi.org/
10.1666/pbio05052.s2). A medial epicondylar
process is present and well developed. It is
perforated by a small entepicondylar foramen.
The supinator crest is also well developed and
extends proximally over one-third the length
of the shaft, such that it overlaps the distal re-
gions of the deltopectoral crests, lying caudal
to them. The olecranon fossa is fairly deep, but
apparently did not perforate the humerus, as
it appears to do in Figure 2 (Sydow [1988] re-
ported that the hole is due to breakage.) The
capitulum and trochlea are distinct, with the
capitulum being convex and the trochlear area
being quite concave. The medial ridge of the
                                                     FIGURE 3. Forelimb elements of mesotheriids. A, Right
trochlea is high, forming a well-defined area        humerus of Mesotherium cristatum in anterior view
of articulation with, and buttress for, the cor-     (modified from Gervais 1869). B, Left elbow joint of Ple-
onoid process of the ulna.                           siotypotherium sp. (MNHN-Bol-V-3724) of Casira, Boliv-
                                                     ia, in an oblique anterolateral view. C, D, Line drawings
   The humeri of Plesiotypotherium archirense        of left manus of Trachytherus spegazzinianus (C) and right
(see Villarroel 1974: Text and Fig. 12) and Ple-     manus (shown as left) of Mesotherium (modified and re-
siotypotherium sp. of Casira, Bolivia, resemble      versed from Ameghino, 1891) (D). Scale bar applies to
                                                     C and D. Abbreviations for B: hum, humerus; rad, ra-
that of Trachytherus. The most notable differ-       dius; rad ses, radial sesamoid; uln, ulna. Those for C:
ences are that the humerus of Plesiotypotherium      Cu, cuneiform; Lu, lunar; Mg, magnum; Sc, scaphoid;
is more gracile than the particularly robust         Td, trapezoid; Tm, trapezium; Unc, unciform.

humerus of Trachytherus (UF 91933) and that
the perforation of the olecranon fossa is nat-         Antebrachium. The ulnae of Trachytherus
ural in Plesiotypotherium. Serres (1867) likewise    and Plesiotypotherium sp. are similar (see Figs.
described the humerus of Mesotherium as ro-          1, 2), having unreduced distal ends and a
bust (Fig. 3A), with well-defined deltoid and        well-developed, long (but not excessively so)
pectoral crests extending two-thirds the             olecranon (damaged in MNHN-Bol-V-3724;
length of the humeral shaft (Serres 1867: p. 743     but see Villarroel 1974: Fig. 13). The olecranon
and the unnumbered table, p. 747; Gervais            of Trachytherus (UF 91933; Figs. 1, 2) curves
1869: Plate XXV.3) and as having a well-de-          medially, providing a large surface area for at-
veloped epicondylar region (the DHW from             tachment of carpal and digital flexors (as in
his data is 0.36). The olecranon process of the      extant scratch diggers like Taxidea, Tamandua,
right humerus was perforate (Fig. 3A), but the       and the subterranean pocket gopher Geomys).
left was not (Gervais 1969: Plate XXII.4). Ser-      The shaft is broadly excavated on the lateral
res noted similarities between the humerus of        surface, likely providing space for massive
Mesotherium and those of several sloths (My-         carpal and digital extensor muscles. An ex-
lodon, Scelidotherium, and Megalonyx) and the        cavation for carpal and digital flexors is pres-
beaver (Castor).                                     ent on the proximomedial surface of the shaft,
236                                        SHOCKEY ET AL.

resulting in an I-beam cross-sectional mor-          similar elbow sesamoid only in the pocket go-
phology in this region. The pisiform (fused by       pher Geomys pinetus (UFm 12358), though
matrix to the distal ulna of UF 91933; Fig. 2C)      some fossorial animals (e.g., Dasypus and
is a robust element that provided insertion for      Manis) have a bony articular process of the hu-
the ulnar carpal flexor muscle (compare with         merus (the HuRl discussed and figured by
Taxidea in Hildebrand and Goslow 2001: Fig.          Szalay and Schrenk [1998]) that shields the lat-
25.6).                                               eral surface of the radius.
   Serres (1867) described the antebrachium of          The distal radius of Trachytherus has well-
Mesotherium as consisting of two distinct and        developed and distinct facets for the scaphoid
robust bones that allow liberal movements. He        and lunate. Distinctive grooves for the ten-
also described the ulna of Mesotherium as hav-       dons of the extensors are also present in Trach-
ing a large, curved olecranon, although that         ytherus and Plesiotypotherium (see Villarroel
shown by Gervais (1869: Plate XXV.5) is not as       1974), as well as in Mesotherium (Gervais 1869:
curved as the olecranon of Trachytherus. The         Plate XXV.4).
shaft is similarly excavated, with a ‘‘gutter’’         Manus. The manus of Trachytherus (Fig.
inside and out (‘‘et parcouru en dedans et en de-    2B) is remarkably similar to that of Mesother-
hors par une vaste gouttière’’ p. 745). He noted    ium (Fig. 3C,D; see also Ameghino 1891: Fig.
that the trochlear groove is broad, allowing         10). Both are pentadactyl, but with a reduced
considerable movement at the elbow joint.            first digit. The remaining four digits are sub-
   The proximal end of the radius is sub-rect-       equal in size with the axis of symmetry run-
angular in T. spegazzinianus and Plesiotypo-         ning between the third and fourth digits. The
therium sp. (see also Villarroel 1974: Fig. 13),     second metacarpal (Mc II) has the most prox-
not rounded as in the Deseadan ‘‘notohippid’’        imal origin, which overlies the base of Mc III;
Eurygenium pacegnum of Salla (Shockey 1997a:         this, in turn, overlies the base of Mc IV, which
Fig. 6.2; Shockey 1997b). Thus, we judge the         overlies Mc V. Mc III is the longest of the meta-
forearm of these mesotheriids to have been           carpals, but extends no further distally than
less capable of supination than that of E. pa-       Mc IV (owing to the more proximal position
cegnum.                                              of its base). These central digits, III and IV, are
   While preparing Trachytherus specimen UF          only slightly more robust and longer than dig-
90960, Sydow (1988) discovered a large sesa-         its II and V. Ungual phalanx III is the best pre-
moid in articulation with the proximal radius.       served and is moderately flattened and bifur-
Such an unusual element is also present and          cated.
fused by matrix to the proximal radius in Ple-          The carpometacarpal joints deviate little
siotypotherium sp. (MNHN-Bol-V-3724; Fig.            from a serial articulation, with most of the in-
3B). Scott (1912: Plate 25.8) reported an elbow      terlocking occurring between the metacar-
sesamoid in the toxodontid notoungulate Ne-          pals. The phalanges are relatively short. The
sodon. It clearly articulated with the proxi-        proximal phalanges have a small palmar
molateral surface of the radius, leaving a dis-      notch for articulation with the poorly devel-
tinctive facet just proximal to the bicipital tu-    oped metacarpal keels. The interphalangeal
bercle of the radius in Trachytherus and Plesi-      joints are simple, lacking palmar notches or
otypotherium, as well as Nesodon (Scott 1912;        the dorsal ‘‘stops’’ seen in many extant dig-
see also Croft et al. 2004: Text and Fig. 6). Pre-   gers (Hildebrand 1985).
sumably, this elbow sesamoid developed in               The ungual phalanges are distinctive, es-
the tendon of the wrist and digit extensors,         pecially (or best preserved) in Mesotherium;
originating on the well-developed distolateral       the distal ends are fissured as in many extant
side of the humerus. The function of such a          fossorial animals (e.g., moles, golden moles,
sesamoid of the elbow is unknown to us, but          and pangolins [see Hildebrand 1985]) and as
it likely helped stabilize the joint and reduce      in some Tertiary notoungulates (Homalodo-
the chance of dislocation. It might also have        therium [see Scott 1912]; Scarrittia [Chaffee
reduced wear on the tendons themselves.              1952]; Eurygenium [Shockey 1997b]). These
Among extant animals, we have observed a             phalanges are blunt, like the distal phalanges
MESOTHERIID MORPHOLOGY                                           237

                                                           moles, marsupial moles, pocket gophers, ant-
                                                           eaters, and armadillos (Hildebrand 1985; Rose
                                                           1999). Sacroischial or other sacropelvic rein-
                                                           forcements are also found in a variety of fos-
                                                           sorial mammals, including moles, pangolins,
                                                           and armadillos (Hildebrand 1985; Rose and
                                                           Emry 1993), as well as pocket gophers. Rose
                                                           and Emry (1993) noted that wombats (Vom-
                                                           batus) and aardvarks (Orycteropus) also have
                                                           sacroischial contact, but that it is not ossified;
                                                           rather, the firm contact is maintained by
                                                           strong ligaments.
                                                              The only pelvis of Trachytherus we are aware
                                                           of is that of UF 90960. It includes all three pel-
                                                           vic bones, but the sacral region is not pre-
                                                           served. The area of attachment of the ilium to
                                                           the sacrum is preserved, but breakage of the
                                                           ischial spine precludes our determination of
                                                           whether this process reached and contacted
                                                           the axial skeleton or only supported connect-
                                                           ing ligaments.
FIGURE 4. Pelves of mesotheriids. A, Pelvis of Plesioty-      Femur. The femur of Trachytherus (UF
potherium sp. (MNHN-Bol-V-3724) of Casira, Bolivia         90960) has a greater trochanter that extends
(left lateral view). B, Pelvis of Mesotherium cristatum
(dorsal view, anterior to left), adapted from Gervais      proximally to about the same level as the head,
1869: Plate XXIV.9. Arrow indicates fusion of ischium to   which lies at the end of an obliquely oriented
the sacral complex. Scale bar applies to both A and B.     neck. The lesser and third trochanters are con-
                                                           spicuous. The center of the third trochanter
                                                           lies about one-third the length down the shaft
of Eurygenium (Shockey 1997b), not clawlike                and terminates at nearly midshaft. The femur
as in Homalodotherium (Scott 1912; Coombs                  of Trachytherus is quite similar to that of Eur-
1983).                                                     ygenium pacegnum (Shockey 1997a: Fig. 6.4;
   Sacropelvic Complex. The pelvic region of               Shockey 1997b), the major difference being
Plesiotypotherium sp. of Casira is most distinc-           that the shaft of Trachytherus is more dorso-
tive in that it preserves, posterior to the ilium,         ventrally flattened in cross-section. The femur
five vertebrae that are solidly fused to one an-           of Plesiotypotherium achirense is also similar to
other (Fig. 4A). (These five do not include the            that of Trachytherus, the most notable differ-
more anterior sacral vertebrae, as they were               ence being the slightly more proximal location
not preserved.) The transverse processes of                of the lesser and third trochanters (see Villar-
the penultimate vertebra contact, and are sol-             roel 1974: Fig. 14). The lesser and third tro-
idly fused to, the ischium. This distinctive               chanters of Mesotherium (Gervais 1869: Plate
morphology is similar to that described by                 XXV.20) are reduced compared to those of
Serres (1867) for Mesotherium (Fig. 4B). Serres            Trachytherus and Plesiotypotherium).
indicated that a total of nine vertebrae are                  Crus. The tibia and fibula are not fused in
fused in the sacral region of Mesotherium and              any of the three mesotheriids studied. The
that the seventh is ‘‘soldered’’ to the ischium            smoothness of the tibiofibular facets suggests
(‘‘on voit l’ischion, se souder d’une manière très-      that movement occurred between the two
intime avec une vertèbre sacrée, la septième.’’ p.      bones, implying that rotation of the upper an-
13).                                                       kle joint occurred.
   Supernumerary fused vertebrae in the sa-                   Pes. Ameghino (1905) described and com-
cral region are found in a variety of fossorial            pared the astragalus of Trachytherus (referred
mammals including talpid moles, golden                     to by its junior synonym, Eutrachytherus) and
238                                                    SHOCKEY ET AL.

                                                                   6.6a), but is fairly short in others (Fig. 5B). The
                                                                   head is subspherical and articulates within a
                                                                   deep concavity of the navicular. In contrast to
                                                                   the condition in many toxodonts (e.g., Eury-
                                                                   genium, Nesodon, Adinotherium) and interath-
                                                                   eriids, both plantar facets lie roughly in the
                                                                   horizontal plane, such that the astragalus
                                                                   overlies the calcaneus (Shockey and Anaya
                                                                   2007).
                                                                      The calcaneum of Trachytherus (Fig. 5A) has
                                                                   a poorly developed, obliquely oriented fibular
                                                                   facet. The ectal facet is more horizontal than
                                                                   that of toxodonts and interatheriids. The lat-
                                                                   eral surface of UF 172514 (Fig. 5A) has a
                                                                   groove for the tendon of the peroneus longus
                                                                   and a distal peroneal process that is not di-
                                                                   rectly adjacent to the distal region of the pe-
                                                                   roneal groove. The apex of the tuber calcanei
                                                                   is rugose and none of the specimens we ex-
                                                                   amined had a well-developed groove for the
FIGURE 5. Photos (above) and line drawings (below) of              Achilles tendon, as occurs in most ungulates
proximal tarsals of Trachytherus spegazzinianus. A, Right          and cursorial carnivores. The cuboid facet is
calcaneum (UF 172514) in lateral (left) and dorsal (right)
views; B, Right astragalus (UF 172437) in dorsal, medial,          slightly concave.
plantar, and distal (clockwise from upper left). The ar-              The transverse ankle joint of mesotheriids
row (plantar view) indicates the groove for the tendon             (i.e., Trachytherus, Plesiotypotherium, and Me-
of the flexor hallucis longus (discussed in text). Scale
bar, 2 cm. Abbreviations: ect f, ectal (lateral) facet; fg,        sotherium) is composed of a ball-and-socket ar-
flexor (flexor hallucis longus) groove; fib f, fibular facet;      ticulation between the astragalus and navicu-
lat tib f, lateral tibial facet; med tib f, medial tibial facet;   lar combined with a modified sliding articu-
nav f, navicular facet; pf, peroneal fossa; pp, peroneal
process of calcaneum; ppa, peroneal process of astrag-             lation of the calcaneocuboid joint. This kind of
alus; sus f, sustentacular facet.                                  transverse ankle joint permits not only exten-
                                                                   sion-flexion but also a considerable degree of
                                                                   supination-pronation of the pes.
that of Mesotherium (Ameghino 1905: Figs. 72–                         From the well-developed flexor groove of
74). He noted these were remarkably similar                        the astragalus of Trachytherus and Mesotherium,
to the astragalus of Orycteropus, except that                      Ameghino (1905) correctly predicted the pres-
the trochlear foramen was absent in the me-                        ence of a great toe in Mesotherium (confirmed
sotheriid specimens (Ameghino 1906). Villar-                       by him [Ameghino 1906]) and Trachytherus
roel (1974: Fig. 16) figured an astragalus from                    (confirmed by a recent discovery of a speci-
Plesiotypotherium that is also very similar to                     men of Trachytherus [MUSM 668] in the late
that of Trachytherus (Fig. 5) and Mesotherium.                     Oligocene [Deseadan] of Moquegua, Peru
The three nearly identical mesotheriid astra-                      [Shockey et al. 2006]).
gali are distinctive in their asymmetric troch-
lear keels; the lateral keel is much larger than                                        Results
the medial and approaches the extreme asym-
                                                                   Paradigm Analysis
metry seen in some extinct sloths (e.g., Mylo-
don, Megatherium; see Owen, 1840). The troch-                         All three mesotheriids (Trachytherus, Plesi-
lear groove is shallow and a separate groove                       otypotherium, and Mesotherium) showed dis-
is present for the passage of the tendon of the                    tinctive morphological characters consistent
flexor hallucis longus, visible in ventral view                    with fossorial habits in all six categories of our
(Fig. 5). The neck is constricted and elongated                    version of Hildebrand’s scratch-digging par-
in some specimens (see Shockey 1997a: Fig.                         adigm (see ‘‘Methods: Paradigm Method’’).
MESOTHERIID MORPHOLOGY                                       239

The functional indices calculated for meso-          sorial mammals, especially the larger diggers
theriids (see Appendix 2) were all within the        such as Orycteropus, Taxidea, and Vombatus.
predicted ranges for scratch diggers.                Such characteristics are qualitatively evident
   Evidence for high out-forces (Fo) for humer-      in the forelimb (Fig. 1) and in the manus, pel-
al retraction at the shoulder joint (Part 1 of the   vis, and hind limbs. Some discrete features re-
paradigm) include the relatively high spine          lated to digging that are found in mesother-
and elongated acromion of the scapula and ex-        iids and various extant diggers include the fis-
ceedingly well developed humeral deltopec-           sured ungual phalanges (as seen in pangolins
toral crests. The distal extension of the delto-     [Manis]), elbow sesamoids (as in pocket go-
pectoral crests provides a long in-lever (Li) re-    phers [Geomys]), and supernumerary ‘‘sa-
sulting in high Fo for humeral protraction (the      crals’’ and fusion between the ischium and the
deltopectoral crests extend 54–71% the length        sacrum (as in dasypodids, myrmecophagids,
of the humeral shaft in Trachytherus and 67%         and Geomys) .
in Mesotherium).                                        Mesotheriids, however, do not compare fa-
   Evidence for high Fo for humeral protrac-         vorably with agile, fast-moving extant mam-
tion (Part 2) includes the tall greater tubercle     mals such as cursorial carnivores and ungu-
of the humerus and the high scapular spine.          lates; they lack the elongated distal limb ele-
   High Fo at the elbow joint (Part 3) is seen in    ments (e.g., metapodials, ulna, radius, tibia)
the moderately elongated olecranon (OL: 37–          typical of such taxa. Additionally, mesother-
46) with Fo increased by the somewhat short          iids exhibit no reduction in digit numbers or
forearm (BI: 90–95).                                 fusion of metapodials, unlike extant cursorial
   High Fo for wrist and digit flexion (Part 4)      perissodactyls and artiodactyls.
is evident in a modestly broad distal humerus           Besides the discrete morphological charac-
(HW: 32–38), medially inflected olecranon,           ters that mesotheriids share with extant dig-
and well-developed pisiform.                         gers, they also closely resemble modern dig-
   Evidence of resistance to forces of the sub-      gers in the proportions of their skeleton. This
strate acting upon the animal (Part 5) is sug-       resemblance is evident in the multivariate
gested in the fissured ungual phalanges of the       analyses comparing mesotheriids with vari-
forelimb. No mesotheriid, however, has the           ous extant taxa of known function.
‘‘bony stops’’ seen in the phalanges of some            Principal Component Analysis. The first
scratch diggers (e.g., Dasypus).                     principal component (PC-1) explained most of
   Hind-limb support for resistance to the           the variation in the data (86.69%), but there
forces of the substrate acting upon the animal       was little variation among the eigenvector co-
(Part 6) is remarkable; Plesiotypotherium and        efficients of the nine variables (Table 1). All
Mesotherium have supernumerary fused ‘‘sa-           were positive and ranged from 0.827 to 0.987.
cral’’ vertebrae and the ischium is strongly         Thus, PC-1 is heavily influenced by body size,
fused to the axial skeleton. Hind-limb               larger animals having higher scores than
strength is also suggested by relatively low CI      smaller ones (Fig. 6A).
and MtF values (CI ⫽ 93 in Trachytherus and             Though PC-2 explained much less of the
83 in Mesotherium; MtF ⫽ 33 in Trachytherus          overall variation (7.69%), there was consider-
and 25 in Mesotherium). Our digging para-            able variation among the eigenvector coeffi-
digm did not predict the mobile ankle joint          cients of the nine variables. The width of the
seen in these mesotheriids, but we note that         distal humerus and length of the deltopectoral
this morphology is similar to that of the fos-       crests had the highest eigenvector loadings
sorial aardvark.                                     (0.456 and 0.333, respectively) and the length
                                                     of the olecranon had a modest influence on
Analog Analysis                                      PC-2 (eigenvector coefficient ⫽ 0.164). These
  Qualitative Results. The three mesotheriids        three morphological variables are associated
examined (Trachytherus, Plesiotypotherium, and       with forelimb strength, thus high positive
Mesotherium) have numerous morphological             PC-2 scores are abstracted as high forelimb
characteristics seen in a variety of extant fos-     out-forces. Conversely, the most negative
240                                                 SHOCKEY ET AL.

TABLE 1.   Summary statistics of multivariate analyses. (PCA and DFA scores for individual taxa given in Appendix
3).

   Summary statistics for Principal Component Analysis of variables (log transformed [ln]) used in the study.
                                            PC-1                         PC-2                          PC-3
Eigenvalues                                  7.802                        0.692                         0.281
% Variance                                  86.688                        7.693                         3.124
Variables
  Radius (L)                                 0.967                      ⫺0.142                        ⫺0.133
  Humerus (L)                                0.981                       0.033                        ⫺0.107
  Humerus (W)                                0.862                       0.456                          0.059
  Deltopectoral (L)                          0.911                       0.333                        ⫺0.161
  Olecranon(L)                               0.882                       0.164                          0.427
  Ulna (L)                                   0.976                      ⫺0.131                        ⫺0.125
  Mt III (L)                                 0.827                      ⫺0.513                          0.155
  Femur (L)                                  0.987                       0.021                        ⫺0.033
  Tibia (L)                                  0.972                      ⫺0.210                        ⫺0.026
           Summary statistics for Discriminant Function Analysis of the nine variables used in the study.
                                             DF-1                         DF-2                         DF-3
Eigenvalues                                   8.838                        0.555                        0.430
% Variance                                   90.0                          5.6                          4.4
Variables
  Radius (L)                                ⫺0.617                       ⫺1.951                        0.668
  Humerus (L)                               ⫺1.615                       ⫺1.622                        3.608
  Humerus (W)                                1.813                         1.152                      ⫺1.529
  Deltopectoral (L)                          1.503                       ⫺1.897                       ⫺0.577
  Olecranon(L)                               1.574                       ⫺0.662                        1.453
  Ulna (L)                                  ⫺1.597                        3.378                        0.609
  Mt III (L)                                 0.146                        1.356                       ⫺0.367
  Femur (L)                                  1.219                        5.472                       ⫺0.316
  Tibia (L)                                 ⫺2.262                       ⫺4.695                       ⫺3.355

FIGURE 6. Principal component analysis of nine variables for 30 extant taxa of known function, two mesotheriid
notoungulates (Trachytherus and Mesotherium) and a notohippid notoungulate (Eurygenium). A, Plot of PC-2 versus
PC-1; B, Plot of PC-3 versus PC-1. Symbols for functional groups are as follows: x, generalist; box, fossorial; triangle,
cursorial; circle, semiaquatic; and diamond, notoungulates of unknown function. (See Table 1 for summary statistics
and Appendix 3 for individual PC scores.)
MESOTHERIID MORPHOLOGY                                               241

FIGURE 7. Discriminant function analysis of nine variables for 30 extant taxa of known function, two mesotheriid
notoungulates (Trachytherus and Mesotherium) and a notohippid notoungulate (Eurygenium). A, Plot of DF-2 versus
DF-1. B, Plot of DF-3 versus DF-1. Symbols for functional groups as in Fig. 6. (See Table 1 for summary statistics
and Appendix 3 for individual DF scores.)

PC-2 eigenvector coefficients were found                  extant species analyzed, 27 (87%) were clas-
among distal limb elements (Mt III ⫽⫺0.513;               sified correctly. Two generalists (the opos-
Tibia ⫽⫺0.210; ulna ⫽⫺0.131). Thus, more                  sum, Didelphis, and the ‘‘false paca,’’ Dinomys)
negative PC-2 scores serve as a proxy for high            were misclassified as semiaquatic, and two
out-velocity. All fossorial taxa had positive             fossorial (both species of Marmota) and one
PC-2 scores (low out-velocity) and nearly all             semiaquatic species (the capybara, Hydro-
cursorial taxa had negative scores. Unspecial-            chaeris) were misclassified as generalists (Ap-
ized and semiaquatic taxa had both positive               pendix 3). Scores on Discriminant Function 1
and negative values. All three notoungulates              (DF-1) clearly distinguished cursorial and fos-
had positive PC-2 scores and shared morpho-               sorial taxa (Fig. 7); this function accounted for
metric space with fossorial taxa, though Me-              90.0% of the variation in the data and was
sotherium had a higher PC-1 value (i.e., was              dominated by high negative loadings for elon-
larger) than any of the extant diggers (Fig.              gated limb elements (especially the tibia, but
6A, Appendix 3, http://dx.doi.org/10.1666/                also the humerus and ulnar shafts) and by
pbio05052.s3).                                            positive loadings for characters associated
   PC-3 accounted for only 3.12% of the vari-             with high out-forces (distal width of humerus,
ance. Positive values were influenced most by             length of deltopectoral crest, and olecranon
olecranon length (eigenvector coefficient ⫽               length). Cursorial taxa thus had low scores on
0.427), and a variety of limb elements had just           DF-1, whereas fossorial taxa had high DF-1
weak influence on the negative values (Fig.               scores; generalists and semiaquatic taxa were
6B). Trachytherus and Mesotherium had positive            intermediate.
PC-3 values and shared morphometric space                    There was little discrimination on DF-2
with fossorial taxa, whereas the PC-3 value for           (5.6% of the variation in the data) and relative
Eurygenium was negative. It shared morpho-                lengths of the hind limb had an important ef-
space with diggers and the semi-aquatic cap-              fect on this function; taxa having long tibiae
ybara (Hydrochaeris). Using varimax rotation,             relative to their femora scored negatively,
log-transforming the data, and adding/sub-                whereas the reverse occurred for those with
tracting taxa/variables did little to change the          relatively short tibiae.
overall pattern of the PCA.                                  The DFA classified Trachytherus and Meso-
   Discriminant Function Analysis. Of the 31              therium as fossorial with high posterior prob-
242                                        SHOCKEY ET AL.

abilities (0.980 and 0.995, respectively) but         high Fo required for digging. Thus, the obser-
very low conditional probabilities (0.034 and         vation of fossorial taxa having short Lo and
0.000, respectively); Eurygenium was classified       long Li is consistent with fundamental prin-
as a generalist, also with a relatively high pos-     ciples of lever mechanics.
terior probability (0.800) and a low condition-          Although there is general concordance be-
al probability (0.162). These disparate proba-        tween the real and ideal (i.e., that predicted by
bilities result from the observation that the         form and that observed), there were some in-
mesotheriids were much closer to the digger           stances of discordance (e.g., ‘‘misclassifica-
group centroid than to any other (posterior           tions’’ in the DFA). Some lack of precision in
probability) but that they were positioned            the relationship between form and functional
near the periphery of digger morphospace              classification results partly from having dis-
(conditional probability). In terms of num-           crete categories for continuous phenomena.
bers, the mean squared Mahalanobis distanc-           Perhaps the greatest technical challenge is
es (MD)2 from the group centroid for extant           when there is little morphological difference
diggers was 1.689 (ranging from 0.422 to              between functional categories. For example,
3.267), that of Trachytherus was 8.688, and that      the most common misclassifications in our
of Mesotherium was 36.12. This is easily seen         DFA were due to morphometric similarities
graphically (Fig. 7). Unlike the PCA, the DFA         between generalists and semiaquatic taxa
was not insensitive to changes in parameters;         (e.g., the semiaquatic capybara was misclas-
log-transforming the data or deletion of taxa         sified as a generalist and two generalists were
included resulted in some changes in classi-          misclassified as semiaquatic). This appears
fication.                                             largely to be a function of both generalists and
                                                      semiaquatic taxa having intermediate DF-1
                   Discussion                         scores. The remaining misclassifications of ex-
   Integrated Analysis. A conspicuous result          tant taxa were those of two species of fossorial
of the multivariate analyses is that the skeletal     groundhogs (Marmota caudata and M. flaviven-
morphology associated with digging is at the          tris). They were misclassified as generalists.
opposite end of a continuum with that asso-           These relatively small diggers were near the
ciated with running (see also Elissamburu             periphery of fossorial morphospace where
and Vizcaı́no 2004). This great difference is         small fossorial taxa approached or shared
consistent with fundamentals of mechanics.            morphospace with generalists. Size seems to
   Whereas high out-forces (Fo) are required          matter, at least in regard to digging: larger
for digging, high velocity defines running. In        diggers had more extreme modifications for
terms of the lever mechanics involved, velocity       digging than the smaller ones, which were
is inversely related to strength (high Fo). This      more similar to generalists.
is best illustrated by comparing the equation            Most important for our immediate purpos-
for Fo with the following equation regarding          es, however, is that no non-fossorial taxa were
the relationship of out-velocity to levers and        classified as fossorial by the DFA. Moreover,
in-velocity (see Hildebrand and Goslow 2001           all larger-bodied extant fossorial taxa were
for discussion):                                      correctly classified as diggers, undoubtedly
                                                      owing to their distinctive morphologies.
                 Vo ⫽ Vi Lo /Li                 (2)
                                                      These distinctive morphologies are likely due
where Vo is the out-velocity (the velocity at the     to the great physical requirement for digging
out-lever), Vi is the in-velocity, Li the in-lever,   in large animals (see Woolnough and Steele
and Lo the out-lever.                                 2001 for discussion). Mesotheriid postcranial
  Note that Vo is proportional to Lo /Li , where-     morphology is highly reminiscent of that of
as Fo is proportional to Li /Lo. For example,         large extant diggers and they are classified to-
whereas elongated Lo (e.g., relatively long           gether in the DFA.
ulna, radius, metacarpals) and short Li (e.g.,           Morphology as a Test for Morphology? In his
relatively short olecranon) facilitate velocity,      important uncoupling of the form-follows-
such a lever ratio fundamentally conflicts with       function dogma, Lauder (1995: p. 13) wrote,
You can also read